The main risk is to children and pregnant women associating with the person receiving I-131 treatment. Since I-131 is excreted in bodily fluids and taken up into the thyroid, those susceptible populations are at increased risk for thyroid cancer if they ingest contaminated bodily fluids. The person receiving the treatment is not because the cells that might become malignant are destroyed by the treatment.
I-131 is a major product of nuclear fallout, and is thought to have contributed to significant thyroid cancer in the USA from Nevada weapons testing.
In 1997, NCI conducted a detailed evaluation of dose to the thyroid glands of U.S. residents from I-131 in fallout from tests in Nevada. In a related activity, we evaluated the risks of thyroid cancer from that exposure and estimated that about 49,000 fallout-related cases might occur in the United States, almost all of them among persons who were under age 20 at some time during the period 1951-57, with 95-percent uncertainty limits of 11,300 and 212,000. The estimated risk may be compared with some 400,000 lifetime thyroid cancers expected in the same population in the absence of any fallout exposure.
http://www.americanscientist.org/issues/feature/2006/1/fallout-from-nuclear-weapons-tests-and-cancer-risks/99999
I presume that distance is advised to avoid gamma radiation emitted by the I-131 in the patient's body rather then the beta radiation which would be largely confined to the thyroid gland, and to emphasize a practical distance to minimize the ingestion of any contaminated fluids.
The question then arises as to what is the safe dose of gamma radiation for a person in whom there is no therapeutic benefit. The answer has to be extrapolated from biophysics and the effect of higher doses since looking at low doses in a population requires too large a number of subjects.
In summary, given our current state of knowledge, the most reasonable assumption is that the cancer risks from low doses of x- or γ-rays decrease linearly with decreasing dose. In light of the evidence for downwardly curving dose responses (see Figs. 2 and 4), this linear assumption is not necessarily the most conservative approach, as sometimes has been suggested (63, 64), and it is likely that it will result in an underestimate of some radiation risks and an overestimate of others. Given that it is supported by experimentally grounded, quantifiable, biophysical arguments, a linear extrapolation of cancer risks from intermediate to very low doses currently appears to be the most appropriate methodology.
http://www.pnas.org/content/100/24/13761.full
Since these are assumptions, the precautionary principle applies. Since the gamma radiation from the body of the treated patient can travel several meters through air, and beta radiation from excreted I-131 1.65 m, then restricting the contact to outside these areas would limit unnecessary exposure.